Title: An intelligent method for human activity recognition based on feature fusion of point clouds from FMCW millimetre-wave radar

Authors: Yingbo Wu; Dunyu Zhao

Addresses: Central China Technology Development of Electric Power Co. LTD., Wuhan, Hubei, 430074, China ' School of Mechanical Engineering and Electronic Information, China University of Geosciences, Wuhan, Hubei, 430074, China

Abstract: To achieve efficient and accurate recognition of human activities, this paper proposes a human activity recognition (HAR) method based on point cloud feature fusion of millimeter-wave radar. To fully extract the features of the activity targets, a point-cloud feature fusion model is designed that combines convolutional neural network (CNN) and attention-based dynamic graph convolutional neural network (DGCNN). CNN captures Doppler features from the Micro-Doppler Spectrogram, while the DGCNN extracts key information from the spatio-temporal features of the point cloud. The combination of these two networks enables activity recognition. Experiments have validated the effectiveness and accuracy of the recognition method.

Keywords: millimeter-wave radar; radar signal processing; activity recognition; neural network.

DOI: 10.1504/IJCSM.2025.151308

International Journal of Computing Science and Mathematics, 2025 Vol.22 No.4, pp.362 - 374

Received: 06 Jun 2025
Accepted: 28 Sep 2025
Published online: 22 Jan 2026 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article